[BUG] Retention lease fails even when the index is active for replication

[skumarp7]

What is the bug?
I have a scenario in ccr where the retention lease is getting expired:

1. Install leader and follower.
2. Set the index settings as such:

"index.plugins.replication.translog.retention_lease.pruning.enabled": "true",
"index.plugins.replication.translog.retention_size": "10 MB", 
"index.soft_deletes.retention_lease.period": "5m"

3. create leader index
4. verify if Follower index is created.
5. Now leave the setup for more than 5m without ingesting new records to leader index.

Expectation:
Index should be in "SYNCING" state.
Result:
Retention lease has been expired for the index, hence the index went to "AutoPaused" state.
What might be the issue? Is this expected? Why did the retention lease expire?

[skumarp7]

Hi @ankitkala , @monusingh-1 ,

Any input ?

[soosinha]

[Triage - attendees 1 2 3 4]

It looks that since the retention_lease.period is quite small and the retention lease renewal might be running after a longer interval due to which retention lease would have expired.
Code reference: https://github.com/opensearch-project/cross-cluster-replication/blob/main/src/main/kotlin/org/opensearch/replication/task/shard/ShardReplicationTask.kt#L280

[skumarp7]

Hi @soosinha ,

Are you saying that the retention lease renewal interval is more than "5m". As per my understanding, doesn't the retention lease gets renewed when the follower polls the leader ? This polling happens even more frequent right ? or is my understanding wrong?

Is there any parameter that controls the retention lease renewal interval ?